A Result on Diffuse Random Measure 1 Introduction

advertisement
Int. J. Contemp. Math. Sci., Vol. 2, 2007, no. 14, 679 - 683
A Result on Diffuse Random Measure
A. Varsei and H. Samimi
Department of Mathematics
Faculty of Sciences, The University of Guilan
P.O. Box 1914 P.C. 41938, Rasht, Iran
varsei@guilan.ac.ir, samimi@guilan.ac.ir
Abstract
We prove a result for a diffuse random measure with Radon intensity,
from which it follows that if φ and ψ are two random measures in Γ
defined on (M(d) , M(d), P ) such that
(d)
(d)
E( φ(I) | MI c )= E( ψ(I) | MI c ) P – a.s.
for all I in the union of a system of a partitions of Γ , then φ = ψ Pa.s.
Keywords: random measure, diffuse, intensity, conditional, Radon
1
Introduction
Let Γ be a locally compact second countable Hausdorff topological space, i.e.
a polish space, B the σ -algebra of Borel subsets of Γ , B0 the family of
bounded sets in B and let M be the space of non negative Borel measures
which are finite on B0 , the space of Radon Measures, and M be the σ algebra in M generated by the mappings μ → μ(B) , B ∈ B0 , (i.e. smallest
σ - algebra making these mappings measurable) and M(d)denotes the subset
of M consisting of all diffuse measures and M(d) = M M(d) . For I ∈ B,
MI denotes the σ - algebra in M generated by mappings μ → μ(B) , B ∈
(d)
B0 , B ⊆ I . MI is defined similarly for any I ∈ B . A random measure
in Γ is a probability measure on (M, M) or a measurable function from some
probability space into (M, M). Diffuse random measure is defined similarly.
If ξ is a random measure, it is clear that for fixed I ∈ B0 , the mapping
ω → ξ(ω, I) is a random variable, which will be denoted by ξ(I) . Also ξ (d) (I)
denotes the restriction of ξ(I) on M(d) .For a Radon measure ω in Γ and any
I ∈ B, define ξ(ω, I) = ω(I) . If P is a probability measure on (M, M), the
measure in Γ given by E( ξ(I) )= ξ(ω, I)P (dω) ,I ∈ B, is called the intensity
of P. Similarly if ξ is a measurable function from some probability space into
680
A. Varsei and H. Samimi
(M, M), the measure in Γ given by E( ξ(I) ) is called the intensity of ξ .
Finally a system of partitions of Γ (compare with null array of partitions in [3]
and see [1]) means a sequence {Δn }n≥1 of partitions of Γ having the following
properties:
1. Each Δn consists of countable many bounded Borel sets.
2. Δn+1 is a refinement of Δn for all n ≥ 1 .
3. For any Γ ⊆ Γ , Γ ∈ B0 , and any ε > 0, there is some n ≥ 1 , such
that Γ can be covered by a finite number of elements of Δn each having
diameter smaller than ε .
The topological structure of Γ ensures the existence of such a sequence of
partitions (see [2]).
In this article we prove that if φ and ψ are two random measure in Γ defined
on
(M(d) ,M(d) ,P ), with Radon intensity, such that
(d)
(d)
(1)
E( φ(I) | McI c )=E( ψ(I) | MI c ) P – a.s.
Δn , where I represents the complement of I , then φ = ψ
for all I ∈
n≥1
P- a.s. (partially case was done in [5] ).
2
Main Theorem
We first represent the following lemmas.
(d)
, M(d), P ) and
Lemma 1 . If φ and ψ are two
random measures in Γ on (M Δn and disjoint I1, I2, ..., Ir ∈
Δn and any
satisfying (1) , for any I ∈
n≥1
a1, a2 , ..., ar ≥ 0 :
n≥1
φ(ω, I)P (dω) =
H
ψ(ω, I)P (dω)
(2)
H
Where H = { ξ(I1) ≤ a1 , . . . , ξ(Ir ) ≤ ar } .
Lemma 2 . Two finite measures μ1 and μ2 on (M(d) , M(d) ) with μ1 (M(d)
)= μ2 ( M(d) ) are identical if, and only if, for any disjoint I1, I2, ..., Ir ∈
Δn
n≥1
r
, ( ξ(I1)1 , . . . , ξ(Ir ) ) maps μ1 and μ2 to the same measure in [ 0, +∞ ) .
Theorem
If φ and ψ are two random measures in Γ defined on (M(d) , M(d) ,P), φ, ψ
have Radon intensity,
(d)
(d)
such that
E( φ(I) | MI c )= E( ψ(I) | MI c ) P – a.s.
Δn , then φ = ψ P- a.s.
for all I ∈
n≥1
681
A result on diffuse random measure
Proof.
Δn , define a measure on M(d) by
For fixed I ∈
n≥1
μ1 (A) = A φ(ω, I)P (dω) , A ∈ M(d) ,
and one by
μ2 (A) = A ψ(ω, I)P (dω) , A ∈ M(d) .
Relation (2), in view of lemma 2 , clearly implies that μ1 = μ2 , i.e.
φ(ω, I)P (dω) =
A
ψ(ω, I)P (dω)
A
for all A ∈ M(d) . This in turn implies that for P- a.e. ω,
Since
φ(ω, I) = ψ(ω, I).
Δn is countable, it follows that there is a P-null event N such
n≥1
that if ω ∈
/ N , then
for all S ∈
φ(ω, S) = ψ(ω, S)
Δn and therefore S ∈ B .
n≥1
The proof of theorem is thus complete.
3
Proof of the lemmas
Proof of lemma 1: (d)
Δn with I ⊆ I0 , and a ≥ 0 , A ∈ MI0c , are fixed.
Suppose I , I0 ∈
n≥1
Define, for n ≥ ν(I) ,( ν(I) denotes the smallest positive integer n such that I
can be expressed as a union of elements of Δn )
Xn (ω) =
χ
(I)
J∈Δn
{ξ(I0 \J)≤a}
(ω)φ(ω, J ).
(I)
Where Δn denotes the family { J ∈ Δn : J ⊆ I } .
For each ω ∈ M(d) , on account of its diffuseness,
lim max ξ(ω, J ) = 0.
n→∞ J∈Δ(I)
n
682
A. Varsei and H. Samimi
Therefore
lim Xn (ω) = χ
n→∞
{ξ(I0 )≤a}
(ω)φ(ω, I).
Since Xn (ω) ≤ φ(ω, I) , from the dominated convergence theorem it follows
that
lim E(χA Xn ) = E(χ Ì
{ξ(I0 )≤a}
A
n→∞
φ(I)).
If we define, for n ≥ ν(I) ,
Yn (ω) =
χ
(I)
J∈Δn
{ξ(I0 \J)≤a}
(ω)ψ(ω, J ),
It follows similarly that
lim E(χA Yn ) = E(χ Ì
A
n→∞
{ξ(I0 )≤a}
ψ(I)).
For each n ≥ ν(I) , in view of (1) we have
E(χA Xn ) = E(χA Yn ).
It therefore follows that
E(χ Ì
A
(3)
Now, if I1 , I2, ..., Ir ∈
{ξ(I0 )≤a}
φ(I)) = E(χ Ì
A
{ξ(I0 )≤a}
ψ(I)).
Δn are disjoint, it is not difficult to see that we
n≥1
(I)
that for each J ∈ Δn0 , either J ⊆ Is for some s, 1 ≤ s ≤ r
can choose n0 such
, or J ⊆ Γ\(I1 ... Ir ) .
Then
φ(ω, I)P (dω)
H
=
Ë Is
(I)
J∈Δn0 ,J⊆
φ(ω, J )P (dω) +
H
(I)
J∈Δn0 ,J⊆Γ\(
Ë Is)
φ(ω, J )P (dω)
H
683
A result on diffuse random measure
=
Ë
(I)
J∈Δ ,J⊆ Is
n0
ψ(ω, J )P (dω) +
H
Ë
(I)
J∈Δ ,J⊆Γ\( Is )
n0
ψ(ω, J )P (dω)
H
from (3) and (1), respectively,
ψ(ω, I)P (dω).
=
H
Proof of lemma 2:
This lemma is a consequence of the following lemma which is consequence
of lemma I.2.1 and Theorem 3.1 of [3].
Lemma 3:
Let φ and ψ be two random measures in Γ (two measurable functions from
some probability space into (M, M)). φ and ψ have
the same distribution in
Δn , the random vectors
M if, and only if, for any disjoint I1, I2, ..., Ir ∈
n≥1
(φ(I1), ..., φ(Ir)) and (ψ(I1), ..., ψ(Ir)) are identically distributed .
References
[1] P. Jacob, P.E. Oliveira, On the conditional intensity of a random measure
35,1 (1994) 103-109.
[2] O. Kallenberg, On conditional intensities of point processes.Z. Wahrsch
verw .Geb.41 (1987) 205-220.
[3] O. Kallenberg, Random measure, Academic Press, 1983.
[4] F. Papangelou., The conditional intensity of general point processes and
an application to line processes, Z. Wahrsch verw.Geb.28(1974) 207-226.
[5] F. Papangelou., Point processes on spacees of flats and other homogeneous Spaces, Math. Proc. Cambridge Pill.Soc. 80 (1976), 297-314.
[6] S.I. Resink, A probability path , Birkhauser Boston, 2001.
Received: August 8, 2006
Download